Loading summary...
7 min read (85% time saved)
Harvey, led by co-founder Gabe Pereyra, is rapidly transforming the legal industry by providing AI solutions to law firms and large in-house teams, moving beyond individual productivity to enhance firm-wide profitability and orchestrate complex legal workflows, while strategically choosing to enable rather than compete with its clients.
Summarize another video
Press ⌘K to quickly paste a new URL
The hosts introduce Gabe Pereyra, co-founder and president of Harvey, setting the stage for a discussion on the company's impact on the legal industry.
Gabe Pereyra introduces Harvey as an AI platform for law firms and large in-house teams, highlighting its rapid growth to nearly 1,000 customers and 500 employees in just over three and a half years. He explains that Harvey is more than a co-pilot; it's an "IDE for lawyers" that connects to vast context and addresses enterprise-level problems like orchestration and governance to make entire law firms more profitable.
Harvey initially targeted large law firms but has expanded to enterprises like Walmart and AT&T, which use the platform for in-house legal work and to collaborate securely with their external law firms. This expansion involves solving technical challenges related to security and data privacy for effective collaboration.
Pereyra clarifies that legal workflows for large firms are far more complex than consumer legal needs. He uses the example of fund formation and M&A in venture capital and private equity, describing the intricate process of structuring entities, drafting agreements, managing diverse investor requirements, and coordinating project management across numerous documents and stakeholders. He likens understanding legal work to deciphering a complex codebase, noting that legal workflows are largely unstructured, similar to programming before advanced AI tools.
The discussion moves to agentic AI, where tasks are broken down into logical steps for AI agents to execute. Pereyra confirms Harvey is implementing this, drawing on his DeepMind background in RL research. He compares associates to agents who receive high-level tasks from partners and perform research, summarization, and drafting. He envisions legal client matters (like fund formation or litigation) as RL environments where models learn to interact with document systems, data rooms, and case law, receiving feedback from partners.
AI will significantly impact law firm structures, particularly leverage ratios and associate training. While AI won't replace partners, it will change how associates are trained and how firms identify future partners. Pereyra draws an analogy to learning programming with AI, where complex concepts can be grasped much faster. He emphasizes that law firms can leverage internal partner feedback and data to train AI. Harvey's goal is to make law firms more profitable by transforming their business holistically, adapting to different practice areas, firm sizes, and client types.
Pereyra elaborates on the analogy between a senior partner and a distinguished engineer, both possessing deep, non-public expertise in architecting complex systems. He cites Gordon Moody's work on Dell's restructuring as an example of a partner's ability to foresee complex legal and financial implications. The challenge for RL in law is the lack of easily verifiable outcomes for long-form text generation, unlike coding with unit tests. Senior partners serve as the 'reward function' by evaluating the quality of legal work. Harvey aims to use internal firm data, including edits and feedback, to train these reward functions, acknowledging this as a significant open research problem.
Harvey has developed a "deployed engineering force" to assist large enterprise clients with customization and integration, a model similar to traditional enterprise software companies like Oracle or IBM. This involves connecting Harvey to clients' business systems (billing, governance) and even building document management systems for those lacking them. This hands-on approach helps Harvey understand client needs and refine its product roadmap. Interestingly, law firms are now starting to offer Harvey implementation services to their in-house clients, creating a new revenue stream.
The rapid adoption of Harvey by law firms, even in its early stages, was surprising but attributed to the transformative nature of AI for text-heavy, knowledge-based industries. The transition from an individual productivity tool to a business transformation platform happened quickly, driven by the inherent value AI brings to legal work.
Responding to a common question, Pereyra explains that building a law firm and a tech company simultaneously is incredibly challenging, as each requires distinct expertise and focus. Harvey's strategic decision is to empower every law firm to become "AI-first" rather than competing with them. This approach allows Harvey to address the larger opportunity of making the entire legal ecosystem more profitable and enabling clients to receive better, faster, and cheaper legal services at scale, avoiding conflicts of interest.
The legal market is vast (trillion-dollar industry), and professional services as a whole are even larger (3-5 trillion). Pereyra highlights the complexity of global transactions, which often require numerous specialized law firms and other professional service providers. Harvey's focus is on building a platform for secure collaboration, data sharing, and AI system deployment across these complex projects, providing infrastructure rather than competing.
Pereyra reflects on his journey from AI research to founding Harvey, noting the significant mental model shift required to scale a company. He emphasizes that Harvey started before GPT-4, driven by a strong conviction in the potential of large language models, fueled by his experience in AI labs and his co-founder Winston's deep understanding of the legal industry. The early bet was that models would continue to improve exponentially, enabling the full complexity of legal work to be addressed. He draws a parallel to coding tools, where the initial form factor for legal AI was clear: document upload and accurate citations.
Harvey is growing rapidly and actively hiring, particularly for strong engineers across various roles, including front-end, product scaling, and AI specialists. They are also scaling their New York office.
Pereyra predicts that people still underestimate the continued exponential improvement of AI models. He foresees a shift from focusing on individual productivity (co-pilots) to organizational productivity, where AI helps large organizations operate more efficiently. He likens this to the Figma transition for designers, where AI will facilitate collaborative work between humans and AI across specialized functions within large companies, optimizing workflows and people management.
The hosts thank Gabe Pereyra for the insightful discussion and provide information on how to follow the podcast and access transcripts.
"At Harvey, we're building AI for law firms and large in-house teams." — Gabe Pereyra
"The big problem we're solving is not how do you make individual lawyers more productive. It's how do you make a team of lawyers working on a client matter more productive?" — Gabe Pereyra
"You can kind of think of associates as agents." — Gabe Pereyra
"I don't think the role of law firm partners actually doesn't change that much in the same way I don't think the role of very senior engineers changes with this because you're largely delegating work." — Gabe Pereyra
"The bigger point is for us it feels like the best outcome is if we can figure out how do we make every law firm how do we help every law firm become an AI first law firm not how do we build one ourselves." — Gabe Pereyra
"I think generally people don't appreciate how much better they're going to continue getting." — Gabe Pereyra
"A lot of people still talk about co-pilots and individual productivity and I think a lot of the things we're starting to think about is like organizational productivity." — Gabe Pereyra
No Priors: AI, Machine Learning, Tech, & Startups
7 min read
Y Combinator
7 min read
Freeplay
7 min read
Weights & Biases
9 min read
No Priors: AI, Machine Learning, Tech, & Startups
9 min read
No Priors: AI, Machine Learning, Tech, & Startups
8 min read
